Researchers have developed a novel framework for assessing the quality of materials data using a confidence-based approach that integrates domain knowledge with graph-structured learning. This method treats data quality as a continuous estimation problem rather than a simple anomaly detection task, allowing for sample-level assessments and comparisons against historical data. The framework incorporates a nine-dimensional feature-quality prior and duplicate-aware sample initialization, calibrated with materials science rules and statistical evidence. Experiments on benchmark datasets demonstrate significant improvements in model performance, with one configuration achieving the highest mean test R² among evaluated screening methods. AI
IMPACT This research could improve the reliability and accuracy of machine learning models in materials science by providing a more robust way to handle imperfect data.
RANK_REASON The item is an academic paper detailing a new methodology for data quality assessment. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- CatalyzeX
- Connected Papers
- cs.LG
- DagsHub
- Frank C. Odds
- Gotit.pub
- HistGradientBoostingRegressor
- Hugging Face
- Litmaps
- Mace
- ScienceCast
- scite Smart Citations
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →